Govern and Optimize MCP Tools with an Agent Gateway

What happens when your AI agent has access to more than 100 MCP tools? Sending every tool definition to the model can consume valuable context and make tool selection less efficient. In this video, we show you how to connect and manage multiple MCP servers with Postman Fabric Gateway. We'll combine 103 tools from a Banking MCP server and a Support MCP server behind a single route, then use progressive tool discovery to reduce the tools initially exposed to the agent from 103 to just three: - `search_tools` finds the most relevant tools using semantic search. - `get_tool` retrieves the schema for the selected tool. - `execute_tool` runs the tool. We'll also walk through configuring upstream credentials, creating routes, inspecting request traces, and the access controls available for governing which users can access specific MCP tools. If you're building AI agents that rely on multiple MCP servers, this approach can help you manage context more efficiently while improving tool discovery, observability, security, and governance. If you find the video helpful, please like it, leave a comment, and subscribe for more videos about APIs, MCP, AI agents, and Postman Fabric Gateway. #MCP #AIAgents #Postman #FabricGateway #APIs